MCP server for LLM prompt-cache observability: prefix diagnostics and per-call white-box report.
Maintain this server? Claim it
This server only runs locally, from a package. Install it with the instructions on this page to try it.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add dobocache \ -- npx -y [email protected]
Environment
DOBOCACHE_LOG_DIR (optional, not in the config above): Directory for the local white-box JSONL log. Absolute path recommended. Optional; defaults to the platform log directory: %LOCALAPPDATA%\DoboCache\logs on Windows, ~/Library/Logs/DoboCache on macOS, $XDG_STATE_HOME/dobocache (fallback ~/.local/state/dobocache) on Linux.DOBOCACHE_WHITEBOX (optional, not in the config above): White-box report card mode: quiet (default, no card in tool output, log only), once (one card next call, then back to quiet), always (card on every call, ~300+ tokens each). Unrecognized values fall back to quiet with a warning. Cumulative stats are always available via cf_report at no token cost.DOBOCACHE_HIT_PRICE_PER_M (optional, not in the config above): Optional price per million tokens for cache-hit tokens used in savings estimates.DOBOCACHE_MISS_PRICE_PER_M (optional, not in the config above): Optional price per million tokens for cache-miss tokens used in savings estimates.We don't run packages, so we can't list this server's tools.